Deepseek Cofounder and Company Information

DeepSeek (Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd.) is a Hangzhou-based startup founded on July 17, 2023. The company describes its mission as pursuing artificial general intelligence (AGI) “that benefits all of humanity,” emphasizing cutting-edge research over immediate commercialization.

DeepSeek is majority‐owned and funded by High-Flyer (Ningbo High-Flyer Quantitative Investment), a successful Chinese hedge fund co-founded in 2016 by Liang Wenfeng (DeepSeek’s founder and CEO). Liang, a Zhejiang University alumnus and quantitative trading pioneer, holds roughly 84% of DeepSeek’s shares. As of early 2025, the company had about 160 employees, many of whom are recent graduates or experts recruited for diverse skills. DeepSeek’s stated focus is purely research — High-Flyer even posted that it will “wholly devote itself to…AI technology that benefits all of humanity” and explore the essence of AGI.

Founders and Leadership

DeepSeek was founded by Liang Wenfeng, who remains its CEO. Liang is the co-founder and controlling shareholder of High-Flyer (an $8–$13 billion AUM quantitative fund), which incubated DeepSeek as a dedicated AI research division. High-Flyer’s other founders (Xu Jin and Zheng Dawei) also share Liang’s computer-engineering background, but media reports focus on Liang as DeepSeek’s public face.

DeepSeek’s flat, lab-like culture is attributed to Liang’s leadership style: colleagues describe a collaborative environment without “996” work hours or rigid hierarchies. In line with its research emphasis, DeepSeek’s staff includes Ph.D. AI researchers and even experts from non-computer-science fields (poets, mathematicians) to broaden its models’ knowledge.

History and Milestones

• July 2023: DeepSeek is founded by Liang Wenfeng under High-Flyer’s umbrella.
• 2024: Released DeepSeek-MoE, DeepSeek-Math, and DeepSeek-V2 models.
• Nov–Dec 2024: Previewed R1-Lite; released DeepSeek-V3 under open source.
• Jan 20, 2025: Launched DeepSeek-R1, a 671B-parameter model with open weights and an MIT license. The mobile app reached #1 in the US iOS App Store.
• Jan 27, 2025: Nvidia’s stock dropped 18% following DeepSeek’s launch; described as a “Sputnik moment” in AI.
• Feb 2025: Competitors respond — ByteDance updates its model; Alibaba announces Qwen 2.5-Max.
• Mar 2025: Baidu releases ERNIE X1 and ERNIE 4.5. DeepSeek releases V3-0324. Government support increases.
• Apr 2025: All major Chinese clouds integrate DeepSeek. Shenzhen announces ¥4.5B fund to support AI. Dozens of mutual funds embed DeepSeek in research workflows.

Investment History and Ownership

• Major Backer: High-Flyer (Liang Wenfeng’s quantitative hedge fund).
• Funding Rounds: None public; no seed/Series funding.
• Capital Raised: Internal (from High-Flyer).
• Valuation: Estimated at $1–10 billion despite limited revenue.
• Ownership: Liang Wenfeng indirectly owns ~84%.

Competitive Positioning

DeepSeek’s R1 model rivals OpenAI’s GPT-4 in performance, but at dramatically lower cost. Training R1 reportedly cost only $5–6 million using ~2,000 Nvidia H800 GPUs. The model uses a sparse Mixture-of-Experts design, activating only ~37B parameters per token to cut compute cost.

DeepSeek is open-source, releasing its models under MIT licenses and publishing full technical reports.

Key Competitors:

• OpenAI: DeepSeek claims parity with GPT-4/o1.
• Baidu: Released ERNIE X1 (rivaling R1) and ERNIE 4.5 in March 2025.
• Alibaba: Announced Qwen 2.5-Max shortly after R1’s release.
• ByteDance: Claims performance surpassing o1 after DeepSeek’s launch.
• 01.AI, Lanmeng, Manus (China): Startups aiming to compete, inspired by DeepSeek’s success.

Financial and Strategic Outlook

DeepSeek’s monetization is minimal. It charges ~$2.19 per million tokens via API — far below OpenAI’s ~$60. Estimated annual revenue per million users is ~$6M.

Its value is in research potential, not revenue. Analysts note DeepSeek’s pricing model and open-source strategy are disruptive, but current income is limited.

Domestically, DeepSeek enjoys significant state and enterprise support:

• Cloud Integration: Baidu, Alibaba, Tencent, Huawei all support DeepSeek models.
• Government Funds: Shenzhen announced ¥4.5B AI compute subsidy.
• Finance Sector: Mutual funds (e.g. China Merchants, E Fund) integrating DeepSeek into research.

Abroad, DeepSeek faces geopolitical tension:

• Singapore Arrests (Feb 2025): Alleged illegal chip exports to DeepSeek.
• US Consideration (Apr 2025): Reports of potential sanctions over access to US AI tech.

Summary

DeepSeek has emerged as a formidable global AI competitor, delivering state-of-the-art open-source models at low cost. Funded by a Chinese hedge fund and driven by research-first philosophy, it has forced global incumbents to respond.

With strong state support and increasing enterprise adoption, DeepSeek’s future hinges on whether it can translate research success into commercial scale amid tightening geopolitical constraints.

DeepSeek’s chatbot topped US App Store charts after its January 2025 launch.

Chinese clouds like Alibaba now offer DeepSeek models for enterprise deployment.